Covary-V3 brings cloud-powered whole-genome intelligence to everyone. Explore now →
Help Center

Learn Covary from first run to shared result.

Practical guidance for preparing sequence data, launching analyses, reading results, downloading raw files, using Browser Visualizer, sharing experiments, and resolving common issues.

Guide

What Covary does and how to think about the results

Covary is designed for alignment-free, translation-aware comparison of DNA sequence records. It helps users explore relationships among sequences through embeddings, distances, dendrograms, and downloadable result files.

Core idea: Covary compares sequence records as encoded biological patterns, then summarizes how records relate to each other in lower-dimensional spaces and hierarchical structures. Outputs are exploratory and should be interpreted with biological context and independent validation.

1. Input

Covary accepts multi-FASTA input. Each FASTA record should have one header and one DNA sequence. Headers become the labels used in plots, tables, dendrograms, downloaded TSV files, and shared result pages.

  • Use simple, meaningful headers such as >speciesA_geneX_sample01.
  • Keep records biologically comparable when possible, such as the same marker, gene, locus, region, or genome scope.
  • Avoid hidden web-page downloads. URL analysis should point to raw FASTA text.

2. Analysis

Covary cleans and encodes records, generates machine-learning-derived embeddings, reduces the representation into PCA, t-SNE, and UMAP spaces, calculates pairwise distances, and reconstructs hierarchical relationships.

PCALinear projection useful for broad structure.
t-SNELocal-neighborhood view useful for clustering patterns.
UMAPManifold projection balancing local and global structure.
DendrogramsHierarchical summaries based on the selected projection.

3. Results

The Results viewer shows embeddings, dendrograms, sequence statistics, analytics, raw downloads, experiment context, sharing controls, and coalescence analysis when the available result payload supports it.

Very large outputs may be capped in the server-served viewer to protect browser and server memory. When a section is capped, download the raw TSV files and use Browser Visualizer to explore them locally.

4. Browser Visualizer

Browser Visualizer is a Pro feature for local-only exploration of downloaded TSV files. Files are loaded in your browser and are not sent back to the web server or compute resources.

  • Use embedding TSVs to render PCA, t-SNE, or UMAP scatter plots.
  • Use distance TSVs to render heatmaps, distance summaries, analytics, and coalescence outputs.
  • Use it when server-served heatmaps or analytics are capped because the run is large.

5. Coalescence Analysis

Coalescence Analysis reconstructs sample-level closeness from repeated label patterns. For headers like marker1_sampleA and marker2_sampleA, Covary can group by marker and sample to infer how samples relate cumulatively across markers.

This does not re-read the original DNA sequence; it uses Covary-derived distances among labeled records. It is useful when several markers, loci, or partitions represent the same sample set.

6. Sharing and context

Shared results are easier to understand when the experiment context is filled in. Add objectives, background, sample information, preprocessing notes, and interpretation notes before sending a public link to collaborators.

For Free and Institutional users, new analyses are public by default unless changed. Pro runs are private by default.

Instructions

Step-by-step workflow

Use this checklist to prepare data, submit a job, interpret results, and share or export outputs responsibly.

01

Prepare a clean multi-FASTA file

Each record should begin with >, followed by a compact label. Use one biological sequence per record. Keep labels unique, readable, and consistent.

>label1_sample1
AAATCTCTATC
>label1_sample2
AAATCTCTATCGGGG
>label2_sample1
AAATCTCTATC
  • Use only DNA characters where possible: A, T, C, G.
  • Decide whether ambiguous characters should be excluded or converted depending on the option available in the run form.
  • Remove hidden spaces, pasted HTML, copied table markup, and non-sequence notes.
02

Choose the input method allowed by your tier

Users may upload a FASTA file. Pro and Institutional users may also use raw URL linking when available. URL linking is useful for cloud-hosted files and collaborative workflows.

  • Upload: best for small-to-medium files and local data.
  • Raw URL: best for cloud-hosted FASTA files that can be accessed as plain text.
  • External storage: Pro users may route future result files to their own supported repository.
03

Start a new analysis

Open the dashboard, enter a project title, choose upload or URL input, select how to handle non-ATCG characters, set public/private status when allowed, then submit the job.

The job table shows queue, running, completed, or error states. When several jobs are waiting, priority and available compute resources determine when a job is dispatched.

04

Read the main result views

EmbeddingsUse PCA, t-SNE, and UMAP views to see group structure.
DendrogramsUse hierarchical clustering to compare relationships.
HeatmapsUse pairwise distances to inspect similarity patterns.
AnalyticsUse summary metrics to identify separations, outliers, and distance ranges.
05

Download raw outputs

Use Raw Files and Raw Distances to download TSV outputs. These files are important for independent review, reproducibility, downstream analysis, and Browser Visualizer.

  • pca_embeddings.tsv, tsne_embeddings.tsv, umap_embeddings.tsv contain plotting coordinates.
  • pca_distances.tsv, tsne_distances.tsv, umap_distances.tsv contain pairwise distance matrices.
  • linkage_*.tsv files contain clustering linkage records.
06

Add experiment context before sharing

In the Results section, add objectives, background, sample information, methods notes, and interpretation notes. This helps readers understand what the dataset represents and what the results should or should not imply.

07

Use the AI-ready prompt carefully

The AI-ready prompt helps users ask another chatbot to explain a Covary result. Remove sensitive sample names before pasting into third-party tools. For free chatbot plans, provide compact summaries first and paste only one small TSV excerpt at a time.

FAQs

Frequently asked questions

Answers to common questions about input files, tiers, privacy, result limits, sharing, storage, and interpretation.

What kind of file should I upload?

Upload a multi-FASTA file containing DNA sequences. Each record should begin with a header line starting with >, followed by sequence lines. Avoid spreadsheet exports, HTML preview pages, or copied text with hidden formatting.

Can I analyze any DNA marker?

Covary can process any DNA sequence collection that fits the accepted input and size limits. Results are most meaningful when records are biologically comparable, such as the same marker, gene, locus, region, amplicon, or comparable genome set.

What does “alignment-free” mean here?

Covary does not require a traditional multiple sequence alignment before analysis. It encodes sequence patterns into features, then uses machine-learning-derived embeddings and distance summaries to explore relationships.

Why are heatmaps or analytics sometimes capped?

Large pairwise distance matrices can become too heavy for server-served visualization and browser rendering. When a cap is reached, download the raw TSV files and use Browser Visualizer locally if your plan allows it.

What is Browser Visualizer for?

Browser Visualizer loads downloaded Covary TSV files directly in your browser. It is useful for large heatmaps, sequence stats, analytics reports, and coalescence analysis when the Results viewer caps server-served visualizations.

Are uploaded Browser Visualizer files sent to Covary?

No. Browser Visualizer reads files locally in the browser. The files are not uploaded to the web server, compute resources, or result storage.

What is Coalescence Analysis?

Coalescence Analysis estimates sample-level closeness from labels that encode marker and sample identity. For example, label1_sample1 and label2_sample1 can be interpreted as two marker-level records from the same sample. The analysis uses Covary-derived distances, not a new sequence alignment.

Why does my result look different across PCA, t-SNE, and UMAP?

Each method emphasizes structure differently. PCA is linear, t-SNE emphasizes local neighborhoods, and UMAP attempts to preserve manifold relationships. Interpret agreement across methods more cautiously than a single projection alone.

Are public links searchable by everyone?

Shared public results can be opened by anyone with the link and may also appear in shared experiment displays depending on visibility settings. If a result should not be public, make it private when your tier allows it.

What does Pro change?

Pro adds private-by-default runs, priority queue behavior, higher runtime allowance, Browser Visualizer access, URL linking where available, and custom data storage routing for future outputs.

Does Pro guarantee that every job will finish?

No. Pro expands functionality, runtime allowance, and queue priority, but it does not guarantee that every dataset is suitable, that every run will finish, or that compute resources will be error-free.

Can I cite Covary?

Yes. Use the Cite Covary page in the app for formatted citations and BibTeX. Include the DOI when referencing the Covary preprint.

Troubleshooting

Common problems and how to resolve them

Most issues are caused by malformed FASTA input, preview-page URLs, capacity limits, capped visualizations, or external-storage configuration.

My FASTA upload is rejected

Confirm the file starts with a > header and contains plain text DNA records. If the file came from a spreadsheet, export or copy it as plain FASTA, not CSV or table text.

  • Remove blank table columns and hidden formatting.
  • Check that sequence lines contain expected characters.
  • Try a very small test FASTA to confirm the workflow.

My URL fails

The URL must point to raw FASTA text, not a preview or sharing page. Open the URL in a private browser window. If you see website buttons, toolbars, or a file preview, it is not a raw file link.

My job stays queued

Queued jobs are waiting for available compute resources. Pro jobs receive priority dispatch. If resources are fully loaded, the dashboard will continue checking and dispatch the job when capacity opens.

My run stopped after a long time

Some tiers have runtime limits. Very large datasets may also exceed practical memory or runtime capacity. Reduce dataset size, use cleaner comparable records, or request Pro access if your use case requires longer runtime.

Heatmaps, stats, analytics, or coalescence are missing

The result may have exceeded the server-served visualization cap. Download the raw distance TSVs and use Browser Visualizer to explore the output locally.

TSV download fails

Confirm the run completed successfully and that result files were stored. If the result was routed to your external storage, download the files from your own repository and use Browser Visualizer.

Browser Visualizer is blank or slow

Large distance matrices can still be heavy for your device. Try a smaller distance file, close other tabs, use a desktop browser, or downsample your dataset before rerunning.

My shared link says unavailable

Confirm the run is public, completed, and not hidden. If result files are stored externally, the shared page may not be able to render the result directly and may ask readers to use downloaded files.

My Pro features do not appear after payment

Refresh the dashboard or sign out and back in. If the account still does not show Pro limits, contact support with the payment time, email used, and quote reference if available.

I need help interpreting a result

Add experiment context, download relevant TSVs, and use the AI-ready prompt as a starting point. For formal scientific conclusions, combine Covary output with metadata, domain expertise, and independent validation.

Still stuck? Contact support with your project title, job status, approximate file size, input method, and a short description of what you expected versus what happened. Contact support →